Skip Navigation LinksHome > May/June 2011 - Volume 60 - Issue 3 > Adolescent Condom Use Consistency Over Time: Global Versus P...
Nursing Research:
doi: 10.1097/NNR.0b013e318217145c
Features

Adolescent Condom Use Consistency Over Time: Global Versus Partner-Specific Measures

Bearinger, Linda H.; Sieving, Renee E.; Duke, Naomi N.; McMorris, Barbara J.; Stoddard, Sarah; Pettingell, Sandra L.

Free Access
Article Outline
Collapse Box

Author Information

Linda H. Bearinger, PhD, MS, FAAN, FSAHM, is Professor, School of Nursing and Department of Pediatrics, Medical School, and Director, Center for Adolescent Nursing, School of Nursing, University of Minnesota, Minneapolis.

Renee E. Sieving, PhD, RN, FSAHM, is Associate Professor, School of Nursing and Department of Pediatrics, Medical School, and Deputy Director, Healthy Youth Development Prevention Research Center, University of Minnesota, Minneapolis.

Naomi N. Duke, MD, MPH, is Adolescent Health Protection Postdoctoral Research Fellow, Center for Adolescent Nursing, School of Nursing, University of Minnesota, Minneapolis.

Barbara J. McMorris, PhD, is Senior Research Associate, Center for Adolescent Nursing, School of Nursing, and Healthy Youth Development Prevention Research Center, Department of Pediatrics, Medical School, University of Minnesota, Minneapolis.

Sarah Stoddard, PhD, MS, is Research Fellow, School of Nursing, University of Michigan, Ann Arbor, Michigan.

Sandra L. Pettingell, PhD, is Research Associate, Center for Adolescent Nursing, School of Nursing, University of Minnesota, Minneapolis.

Accepted for publication February 4, 2011.

This study was supported with funds from the National Institute of Nursing Research of the National Institutes of Health (5R01NR008778; Principal Investigator [PI]: Sieving). In addition, during the preparation of this manuscript, the authors were supported by a nursing training grant (T80-MC00021; Center for Adolescent Nursing, PI: Bearinger) from the Maternal and Child Health Bureau (Title V, Social Security Act) Health Resources and Services Administration, Department of Health and Human Services; a Health Protection Research Initiative Training Grant (T01 CD000185; Adolescent Health Protection Research Training Program, PI: Bearinger); and a Prevention Research Centers Cooperative Agreement (U48 DP001939; Healthy Youth Development Prevention Research Center, PI: M.D. Resnick) from the Centers for Disease Control and Prevention.

The editorial opinions are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention.

Corresponding author: Linda H. Bearinger, PhD, MS, FAAN, FSAHM, Center for Adolescent Nursing, School of Nursing, University of Minnesota, 5140 Weaver-Densford Hall, 308 Harvard Street, S.E., Minneapolis, MN 55455 (beari001@umn.edu).

Collapse Box

Abstract

Background: The conundrum of measuring condom use consistency, particularly with adolescents, has left researchers with a cacophony of strategies, thereby limiting comparability and interpretation.

Objective: The aim of this analysis was to compare and contrast two measures of condom use consistency, global versus partner specific, and their relationships with key covariates, using trajectory groups differentiated by stability of condom use consistency over three time points.

Method: Using self-report data from sexually active girls (aged 13-17 years) in a clinic-based intervention study aimed at lowering risk for early pregnancy, this analysis compared two measures of self-reported condom use consistency: (a) a global measure: overall condom use consistency in the past 6 months and (b) a partner-specific measure: condom use consistency with the most recent sex partner in the last 6 months. Using a subjective rule-based approach, the adolescent girls in the study (n = 151) were classified into trajectory groups representing their condom use consistency at three time points (baseline and 6 and 12 months). Then, using bivariate methods, trajectory groups were compared on four baseline covariates (age, treatment condition, hormonal use in the last 6 months, and number of sex partners in the last 6 months) and three time-varying covariates measured at baseline and at 6 and 12 months (hormonal use stability, stability of primary sex partner, and stability of number of sex partners).

Results: For the trajectory groups formed using the global measure of condom use consistency, stability of the primary sex partner differed significantly between trajectory groups. For the partner-specific trajectory groups, two baseline and one time-varying covariate relationships were significant: hormonal use in the 6 months prior to baseline, number of sex partners in the past 6 months (baseline), and stability of the primary sex partner (time varying), with hormonal use stability (time varying) trending toward significance.

Discussion: The larger number of significant covariate relationships with the partner-specific trajectory groups suggests greater utility in assessing partner-linked behavior rather than a global measure. Despite limitations of the analytic strategy, this study sheds light on a measurement conundrum that has been an obstacle to comparing and contrasting indicators of condom use consistency during adolescence.

Self-reported condom use is a key variable in large-scale surveys of health behaviors and in evaluation studies used to assess the impact of interventions focused on lowering health risk and improving sexual health outcomes. The importance of accurate condom use measurement within adolescent and young adult populations cannot be overstated, as information gleaned becomes a benchmark in setting public health goals, planning and evaluating programs, mobilizing advocacy efforts, and identifying resources for funding. In clinical settings, accurate condom use self-report is critical, as brief office interventions are tailored to information obtained from a young person's behavioral report.

Measurement of condom use encompasses a broad range of possibilities ranging from community- and population-level activity (e.g., examining retail sales) to collecting individual biomarker data (e.g., detection of ejaculate, sexually transmitted infection in vaginal samples; Catania, Gibson, Chitwood, & Coates, 1990; Rose et al., 2009). The most common mode for measuring condom use involves methods in which young people provide direct reports of their condom use through self-administered questionnaires (SAQs), computer-assisted self-interviews (CASIs), daily health diaries, momentary sampling methods, and face-to-face interviews. Although adolescents are able to provide reliable self-reports of condom use and other sexual behaviors (Sieving et al., 2005; Younge et al., 2008), discordances in reporting are noted between modalities (McAuliffe, DiFranceisco, & Reed, 2007; Rose et al., 2009) and there is no consensus on which modality yields the most accurate data (Weinhardt, Forsyth, Carey, Jaworski, & Durant, 1998). Because there is no gold standard with respect to validation of the self-report measurement approach, minimization of measurement error is essential (Catania et al., 1990; Sieving & Shrier, 2009; Younge et al., 2008).

From a measurement error perspective, the validity of adolescents' self-reported condom use may be affected by cognitive factors and by the situation in which assessment takes place. Adolescents' varying cognitive abilities can affect comprehension of self-report questions. In addition, the ability to recall sexual behaviors is affected by the length of the recall period, the frequency and complexity of behaviors, and the personal salience attributed to the behavior (Brener, Billy, & Grady, 2003; Catania et al., 1990; Geary, Tchupo, Johnson, Cheta, & Nyama, 2003). With solitary, infrequent, or salient experiences, such as first sexual intercourse, recall over a long period may be feasible. However, with increasing frequency of sexual activity, number of sexual partners, and variation in the type of sexual encounters, accurate recall of specific behaviors may be more difficult (Brener et al., 2003; Catania et al., 1990). For example, sexually active adolescents tend to be unreliable in their reporting of number of vaginal intercourse episodes over a 6-month period (Sieving et al., 2005).

Young people's tendencies to distort self-report may be influenced by the adolescent respondent, the measurement instrument, and the interview context (Sieving & Shrier, 2009). Adolescents exist in a broad social structure stratified by age, gender, race, and class; these factors may be important mediators of self-presentation bias (Sieving & Shrier, 2009). Age likely affects the self-presentation of behaviors in which social norms differ for younger versus older adolescents. Young people's concerns about investigator expectations, perceptions of peer behaviors, and sensitivity to social cues about acceptable behaviors may also influence self-report (Catania et al., 1990; Geary et al., 2003). Regarding instrumentation, questions about sensitive behaviors present particular challenges to honest self-disclosure and accurate self-presentation. Explaining terminology and framing questions using nonjudgmental and developmentally appropriate language may minimize bias (Catania, 1999; Sieving et al., 2005). Interview strategies, for example, perceived lack of privacy, confidentiality, or anonymity, can pose substantial threats to honest self-disclosure. Evidence suggests that bias can be minimized using an interview mode, such as SAQ and CASI, that allows for more privacy than face-to-face interviewing (Brener et al., 2003; Catania et al., 1990; Sieving & Shrier, 2009). In sum, factors related to adolescents' cognition, study instrumentation, and data collection methods could result in underreporting or overreporting of sexual behaviors with subsequent misrepresentation of prevalence or change in these behaviors (Catania et al., 1990).

From a developmental perspective, a number of individual (e.g., self-efficacy, values, and attitudes), relational (e.g., relationship duration, relationship type, and partner communication), and contextual (e.g., access to condoms, adoption of hormonal contraceptive, and substance use) factors may impact sexual behaviors and ultimately confound assessments of behaviors such as condom use. In a nationally representative sample of adolescents, Manlove, Ryan, and Franzetta (2007) identified several correlates of more consistent contraceptive use: prior contraceptive use, greater partner homogamy, more intimate and couple-like activities within a relationship, and better communication about sex with a partner. Perceptions of being in a romantic relationship have been negatively associated with contraceptive use consistency (Katz, Fortenberry, Zimet, Blythe, & Orr, 2001; Manlove et al., 2007). A more recent study found an array of relationship characteristics to be negatively associated with consistent condom use: controlling partner behavior, mistrust, jealousy, perceived partner inferiority, enmeshment, and perceived relationship importance in the context of increasing relationship duration (Manning, Flanigan, Giordano, & Longmore, 2009). Among adolescent women in clinic settings, condom use was found to be more likely in new versus established relationships (Fortenberry, Wanzhu, Harezlak, Katz, & Orr, 2002; Wiemann et al., 2009). For those in first sexual relationships, having an older partner and having a longer relationship have been associated with interruptions in contraceptive use (Manlove & Terry-Humen, 2007). Assessing only behaviors with a primary sex partner, Ott, Adler, Millstein, Tschann, and Ellen (2002) found female adolescents' condom use to be negatively correlated with hormonal contraceptive use and increasing age. In contrast, Crosby et al. (2007) did not see a reduction in condom use among adolescent girls using hormonal contraception.

As noted earlier, self-reported condom use can be obtained through different modalities. Self-administered questionnaire, the most common modality, is less expensive to use and may not feel as intrusive as face-to-face interviews (Catania et al., 1990; Sieving & Shrier, 2009). However, data quality may be limited by respondent literacy and investigator adeptness in using understandable terminology (Catania et al., 1990). Moreover, SAQ does not allow for follow-up on confusing responses (Catania et al., 1990; Weinhardt et al., 1998). Computer-assisted self-interview, a now widely adopted survey method, increases privacy and allows for the use of branching; CASI with audio (audio-CASI, or A-CASI) alleviates limitations of respondent literacy (Couper, Tourangeau, & Marvin, 2009). Diaries permit collection of information over shortened time intervals, for example, for each 24-hour period, and behaviors may be recorded within hours of occurrence. Consequently, recall bias may be reduced with greater accuracy in the reported frequency and temporal sequence of specific behaviors while also documenting greater detail regarding contexts for sexual behaviors (Catania et al., 1990; Katz et al., 2001). Keeping daily diaries requires effort and could involve recording sensitive information. Therefore, diary study participants may self-select and not represent a population of interest. Sexual behaviors may be reported with greater (Ramjee, Weber, & Morar, 1999) or lesser (Coxon, 1999) frequency on diaries versus surveys. Respondents may fatigue from daily reporting, making inaccurate entries (Ramjee et al., 1999). Face-to-face interviews can minimize nonresponse and permit further probing of confusing or complex answers, correcting of misperceptions, and clarifying of meaning (Catania et al., 1990; Weinhardt et al., 1998). However, face-to-face interviews are more costly and less efficient and may be influenced by respondent comfort or perceptions of interviewer characteristics (Catania et al., 1990; Weinhardt et al., 1998). Momentary sampling methods (e.g., cellular phones and handheld electronic devices), by permitting real-time assessment of behaviors within a specific context, have significant advantage over traditional survey methods and daily diaries (Shrier, Shih, & Beardslee, 2005). However, momentary sampling has disadvantages. Frequent data collection increases respondent burden, which may adversely affect retention and adherence. Moreover, it requires the purchase and maintenance of electronic data collection devices (Sieving & Shrier, 2009).

Within study modalities (e.g., SAQs, CASIs, and face-to-face interviews), assessments of condom use consistency vary. The most commonly used referents are condom use over a given period (e.g., "How often in last [period: 2 weeks, 2 months, 6 months] did you use a condom?"), condom use on specific occasions (e.g., use at first or last intercourse), measures of condom nonuse (e.g., number of times condom is used subtracted from number of intercourse events), proportional measures of consistency of condom use (e.g., number of intercourse events with a condom divided by the number of intercourse events), and composite indices of condom use that combine absolute and proportional measures (Graham, Crosby, Sanders, & Yarber, 2005; Noar, Cole, & Carlyle, 2006). Assessments vary by length of recall period; strategies to improve respondent recall capabilities include use of cues and calendars (Graham et al., 2005).

With increasing recognition that consistency of condom use varies between partners, measuring condom use specific to each sex partner is becoming more common (e.g., condom use in reference to casual vs. steady partners; Graham et al., 2005). In a recent review, Noar et al. (2006) concluded that condom use measurement has improved in a number of ways, including better measurement types and recall periods, greater specificity to sexual acts, and increased assessment of test-retest reliability, social desirability, and condom use skills. However, use of varied methods presents challenges in comparing findings across studies, in deciphering which questions produce the most accurate answers, and in identifying interventions most likely to increase condom use and reduce sexual risk (Noar et al., 2006).

Although a significant body of literature documents varied approaches to assessing condom use, few studies have compared global versus partner-specific measures of condom use consistency and their relative ability to capture their associations with changes in related behaviors, such as changes in hormonal use, primary partner, and number of sex partners, that are so prevalent during adolescence. Using data from a clinic-based youth development intervention study designed to reduce health risk behaviors linked to early pregnancy in a high-risk sample of adolescent girls, the purpose of this study was to model condom use consistency across three waves of data using two self-report measures of condom use consistency. Using global (condom use consistency in the past 6 months) and partner-specific (condom use consistency in the past 6 months with the most recent primary partner) condom measures to create trajectories of condom use consistency, we sought to answer the following question: Is one measure versus the other more associated with variables expected to be related to condom use consistency? In demonstrating the capacity of a particular measure to capture changes in behavior over time, the groundwork is laid for better uniformity of adolescent condom use consistency measurement across observational and intervention studies and in clinical settings.

Back to Top | Article Outline

Methods

Setting and Sample

All the adolescent girls in the sample were enrolled in a clinic-based youth development intervention trial, called Prime Time, aimed at precursors of teenage pregnancy including sexual risk behaviors, violence involvement, and school disconnection (Sieving, Resnick, et al., in press). Prime Time was conducted in two community-based clinics and two school-based clinics in Minneapolis and St. Paul, Minnesota. The multifaceted intervention strategy involved one-on-one case management, peer leadership, and service learning over an 18-month period. Sexually active 13- to 17-year-old girls who met one or more of five risk criteria were invited to participate in the study. These criteria included clinic visit involving a negative pregnancy test, clinic visit involving treatment for sexually transmitted disease, high-risk sexual and contraceptive behaviors, violent behaviors, and/or behaviors indicating school disconnection. Behavioral risk criteria were assessed through a 20-item self-report screening instrument (Sieving, Resnick, et al., in press). Girls who were married, pregnant, or parenting were excluded from this study. Of the 1,434 girls who completed the screening tool, 571 (39%) met eligibility criteria; of these, 253 (44%) agreed to participate and provided written informed consent (Sieving, McMorris, et al., in press). The institutional review boards of the University of Minnesota and participating clinics approved all study protocols.

After consent was obtained, baseline data were gathered using A-CASI methods and the girls were randomized to study conditions; the current analysis used those assigned to both intervention (n = 126) and control (n = 127) groups. At baseline and at 6 and 12 months, girls completed surveys with questions about demographics, sexual behavior, and contraceptive use. Attrition was minimal, with 96% of the eligible sample retained by the 12-month survey (Sieving, McMorris, et al., in press).

To be included in the current analysis, girls had to have completed self-report A-CASI surveys at baseline and at 6 and 12 months, answered items about condom use consistency, and been sexually active at all three time points. A total of 185 girls (73% of the original sample of 253) met these criteria. This initial sample of 185 represented diverse race/ethnicity groups; the highest number (43%) was Black/African/African American, followed by those reporting multiple races/ethnicities (21%), Asian/Asian American/Pacific Islander (14%), White/European American (10%), Hispanic/Latina (10%), or American Indian (2%). At baseline, nearly two-thirds (64%) had lived in the same home for the 6 months prior; almost half (47%) lived with one parent only and 42% lived with both parents. Close to half (47%) indicated that their families were receiving public assistance. Almost one in five (19%) had changed schools within the past year (not related to transitions between elementary, middle, and senior high school), and another 17% had changed school two times or more in the previous year. Only 4% (n = 8) were not in school at baseline.

Back to Top | Article Outline
Measures

All data were provided by answering questions on A-CASI. Research staff trained in A-CASI administration oriented study participants to the computer survey with several nonsensitive trial items. To minimize self-disclosure bias, data were collected in locations offering privacy and research staff routinely assured participants of the confidentiality of their responses. Reliability of each of the condom use consistency measures described below has been detailed elsewhere (Sieving et al., 2005).

Back to Top | Article Outline
Global Condom Use Consistency

Dichotomizing condom use consistency is an approach recommended by Noar et al. (2006) in a review of condom use measurement across 56 studies of sexual risk behavior. In this study, global condom use consistency was a girl's estimate of the frequency with which condoms were used during sexual intercourse with all partners over the 6-month period prior to data collection. All girls were asked: "How often in the past 6 months did you use a condom?" Five response options were used: 1 = never, 2 = less than half the time, 3 = half the time, 4 = more than half the time, and 5 = every time. Responses were dichotomized to reflect little/no use (responses 1-3) versus most/every time use (responses 4 and 5) at baseline and at the 6- and 12-month time points.

Back to Top | Article Outline
Partner-Specific Condom Use Consistency

Respondents also indicated whether they used a condom in the previous 6 months with their most recent partner and then subsequently answered questions about frequency of use with this sexual partner. Information for each of the prior 6 months was collected using a monthly calendar embedded in the A-CASI instrument. For each month, responses were 1 = never, 2 = some of the time, 3 = most of the time, and 4 = every time. The number of months a participant reported using condoms most of the time or every time (responses 3 and 4) she had sex with her most recent partner was counted. The total 6-month count was divided by the number of months the participant reported having sex with this partner, resulting in a U-shaped distribution ranging from 0 to 1, reflecting the proportion of months in which girls used condoms consistently. Using an empirical approach described by Crosby, Yarber, Sanders, and Graham (2004), we compared the small number of girls reporting inconsistent use with their most recent partner (the bend in the U-shaped distribution) with nonusers and consistent users and determined that inconsistent users looked similar to girls who reported no condom use on a range of descriptive variables (i.e., age, race/ethnicity, attitudes toward birth control, number of male sex partners in the past 6 months, length of sexual relationship with the most recent partner, condom use self-efficacy, and sexual communication with partner). Therefore, a dichotomous variable was created in which 1 = consistent condom use with most recent partner (or proportion = 1) and 0 = consistent nonuse of condoms with most recent partner (proportion < 1). Thus, the partner-specific measure was directly comparable with the global condom measure.

In identifying covariates, we first considered group assignment, that is, treatment or control group, and participant age at the time of the baseline survey. Second, we examined relationships with several key variables conceptually related to condom use consistency, as described previously. The set of covariates employed in this study included measures that may associate with condom use consistency at study baseline (referred to as baseline covariates) as well as measures that may predict stability (or change) in condom use consistency over the three survey time points (referred to as time-varying covariates).

Back to Top | Article Outline
Hormonal Use (Baseline and Time Varying)

The first of two measures of hormonal use, baseline hormonal use, was a dichotomous indicator contrasting whether a participant used some method of hormonal birth control (i.e., birth control pills, the Depo-Provera shot, Orthro Evra, or the Nuva-Ring) during the past 6 months as reported on the baseline survey. The second measure, hormonal use stability (a time-varying covariate), represented whether girls used some form of hormonal contraception in the 6 months prior to each of the three survey points.

Back to Top | Article Outline
Number of Sex Partners (Baseline and Time Varying)

For number of sex partners at baseline, girls were asked on the baseline survey how many males they had vaginal sex with during the last 6 months. A second time-varying measure was created to reflect having one partner (albeit not necessarily same partner) at all three survey time points, that is, stability of number of sex partners. Girls who reported having only one male sex partner at each time point were coded as 1; all others were coded as 0.

Back to Top | Article Outline
Stability of Primary Sex Partner (Time Varying)

Stability of primary sex partner was the consistency with which a respondent named the same primary sexual partner at multiple waves of data collection. Respondents provided initials for the first and last names of their primary sexual partner in the previous 6 months, as well as the age of the partner, dates of first and last intercourse, and type of relationship with the partner. Coding of primary partner stability was based on matching this identifying information across periods. Primary partner stability was scored 1 when identifying information for the primary partner matched at all three periods and was scored 0 otherwise.

Back to Top | Article Outline
Data Analysis
Processes for Determining Trajectory Analysis Approach

Trajectory of condom use consistency is the pattern of reported consistency as determined at multiple time points. Two approaches are available for the creation of trajectory groups: statistical modeling and subjective classification rules. Statistical modeling, completed within programs such as SAS or MPlus, uses a formal statistical structure to determine the placement of individuals into groups based on patterns of change over time. To create trajectory groups using subjective classification rules, researchers place individuals into groups based on identifiable, predetermined characteristics. Both methods were considered during the analytic process, and the decision to use subjective classification rules to determine our trajectory groups was based on several factors.

First, the sample size was relatively small, with only three points in time with which to observe change. Second, condom use consistency, for both measures, was conceptualized and operationalized as a dichotomous variable rather than continuous, interval, or ordinal. This strategy captured our conceptualization of condom use consistency as a "type" or group of girls who are consistent versus inconsistent and/or non+users (the creation of these groups is detailed in the section Formation of Trajectory Groups) rather than a continuum of use or a compilation, over time, of a single, repeated behavior. Although software programs have the capability of modeling trajectories of categorical outcomes, the availability of only three time points and a small sample size limited our ability to determine multiple distinct trajectory groups. Only a finite number of trajectories are possible when a dichotomous outcome is measured at three time points (23 = 8 trajectory groups).

One other factor determined our final analytic plan. Initial analysis revealed considerable lack of within-person variability in condom use consistency among our study participants, whether measured globally or specific to the most recent partner. Most likely, this stability in our outcome variables, along with all the factors detailed previously (dichotomous outcomes, small number of time points, and small sample), contributed to model convergence problems when trying growth curve-based statistical approaches to identify latent groups.

Back to Top | Article Outline
Formation of Trajectory Groups

Considering the constraints detailed previously, subjective classification rules were used to create trajectory groups based on self-reported measures of global and partner-specific condom use consistency at three data collection points, that is, baseline and 6 and 12 months. Study participants were classified into five trajectory groups for each of the condom use consistency measures, that is, global and partner specific. With 1 representing condom use every or most of the time and 0 representing no or little use, the following groups were created: stable use (1,1,1; baseline and 6 and 12 months, respectively), stable nonuse (0,0,0), change toward use (0,0,1 or 0,1,1), change toward nonuse (1,0,0 or 1,1,0), and unstable change (1,0,1 or 0,1,0). The number and percentage in each of these five trajectory groups varied, depending on whether the global measure or the partner-specific measure was used to assign participants to groups. Table 1 shows the distribution of participants in the five trajectory groups as a cross-tabulation between the two measures of condom use consistency. For 65% (n = 120) of the girls, trajectory group assignment did not vary by measure used for classifying participants (bolded cells in Table 1; e.g., 26 girls were classified in the stable use trajectory group on both measures). The remaining 65 girls (36%) were assigned to different trajectory groups based on their self-reported condom use consistency for the global measure versus their most recent partner, at any of the three time points.

Table 1
Table 1
Image Tools

As noted in Table 1, when group assignment was determined by the global measure, 13 participants were classified as having unstable change (1,0,1 or 0,1,0) over time; for partner-specific group assignments, 28 showed unstable change over time. Seven participants were classified in the unstable change group for both of the two outcome measures-global and partner specific. However, because the two trajectory patterns defined as unstable (1,0,1 or 0,1,0; V or inverted V for the dichotomous outcome measures-a common distribution for condom use measures as reported by Crosby et al., 2005) represent distinct patterns of behavior, it would be difficult to discern patterns of relationships between the covariates and these unstable change trajectory groups. (Note: Empirically comparing the groups with these two patterns, using a screening analysis laid out by Crosby et al. (2004), it was found that they were too heterogeneous to combine into one group representing unstable change for either variable. In addition, there were too few of these girls, especially as measured by global condom use, to include as separate groups of unstable desirable and unstable undesirable change.) Therefore, this fifth group classified as unstable change in condom use consistency was omitted from the bivariate analyses to eliminate the possibility of confounding relationships. This elimination reduced the final sample to 151 participants.

Back to Top | Article Outline
Final Analytic Plan

The goal was to determine, for which measure of condom use consistency (global vs. partner specific), were covariates better able to distinguish between patterns of change and stability in this behavior over three time points. Tests for significant relationships between baseline and time-varying covariates and condom use consistency trajectory groups were conducted using Pearson's chi-square for cross-tabulations and F test for analyses of variance. SPSS v.14.0 was used for the analysis. Nominal p values of .05 were used.

Back to Top | Article Outline

Results

Descriptive Findings

Table 2 provides descriptive statistics for the covariates for the 151 girls in the final sample. The mean age for participants was 15.7 years (SD = 1.05 years). With random group assignment, equal sized treatment and control groups were expected. Close to two-thirds of the girls used hormones at baseline; however, over time, 61% changed hormonal use status (from use to nonuse or nonuse to use) at each consecutive wave of data collection. In terms of consistency of primary sexual partner at each point in time, two-thirds named a different primary partner in consecutive waves. Although the range of number of sexual partners in the 6 months prior to baseline varied from one to seven (girls reporting no sexual partners were ineligible to participate in the study), the majority (63%) reported only one sexual partner. However, at least once during the three waves of data collection, 60% reported that they had more than one sexual partner in the previous 6 months.

Table 2
Table 2
Image Tools

As noted in Table 3, the largest trajectory group reported "little or no" condom use at all three waves (0,0,0; 36% and 42%, respectively, for global and partner-specific trajectory groups). The second largest trajectory group reported condom use "every/most" of the time at all three waves (27% and 21%, respectively, for global and partner-specific trajectory groups).

Table 3
Table 3
Image Tools
Back to Top | Article Outline
Bivariate Findings

Relationships between condom use consistency trajectory groups and the continuous covariates measured at baseline are presented in Table 3. Relationships between trajectory groups and dichotomous covariates are presented in Figures 1-4.

Figure 1
Figure 1
Image Tools
Symbol 1...
Symbol 1...
Image Tools
Symbol 2...
Symbol 2...
Image Tools
Symbol 3...
Symbol 3...
Image Tools
Figure 2
Figure 2
Image Tools
Figure 3
Figure 3
Image Tools
Figure 4
Figure 4
Image Tools
Back to Top | Article Outline
Global Condom Use Consistency Trajectory Groups

Six of the bivariate relationships for the trajectory groups classified using the global condom use consistency measure were nonsignificant: (1) baseline covariates including age, intervention group, number of sexual partners in the last 6 months (Table 3), and hormonal use in the last 6 months (Figure 1) and (2) time-varying covariates including hormonal use stability (Figure 2) and stability of number of sex partners (Figure 4), that is, having only one sex partner at baseline and at 6 and 12 months. The only significant relationship with global condom use consistency trajectory groups was the time-varying covariate, stability of primary sex partner, that is, having the same primary sexual partner at baseline and at 6 and 12 months (p = .023). Smaller percentages of girls in the stable use and change to use trajectory groups reported the same primary sex partner at the three time points as compared with percentages in the change to nonuse and stable nonuse groups (Figure 3). In other words, staying with the same primary partner over time was significantly associated with trajectories leading to condom nonuse.

Back to Top | Article Outline
Partner-Specific Condom Use Consistency Trajectory Groups

Most bivariate relationships between the covariates and the partner-specific trajectory groups were significant, as shown in Table 3 and Figures 1-4. As anticipated, both age and the intervention group (treatment or control) covariates were nonsignificant. However, three of the covariates expected to be related to condom use consistency were significant: number of sex partners in the past 6 months (reported at baseline; p = .023), baseline hormonal use (p = .034), and the time-varying covariate, stability of primary sex partner (p = .007). An additional covariate trended toward significance: hormonal use stability (p = .088). Specifically, groups reporting consistency of condom use at baseline (stable use and change to nonuse) had the highest number of baseline sexual partners (Table 3; M = 1.69 and 2.0 partners, respectively, vs. M = 1.47 and 1.32 partners for groups reporting nonuse at baseline). In addition, smaller percentages of girls in the stable use and change to use trajectory groups reported using hormonal methods in the 6 months prior to baseline as compared with percentages in the change to nonuse and stable nonuse groups (Figure 1). In addition, a greater percentage of girls in the change to nonuse trajectory group reported hormonal use at each of the three survey points (i.e., hormonal use stability) as compared with other trajectory groups (Figure 2). A smaller percentage of girls in the change to use group reported the same primary sex partner at the three time points as compared with other trajectory groups (Figure 3). The only covariate for which significance was anticipated but not found was having only one sex partner at baseline and at 6 and 12 months (Figure 4).

Back to Top | Article Outline

Discussion

This analysis of two measures for assessing condom use consistency (global and partner specific) among a group of sexually active 13- to 17-year-olds afforded the opportunity to examine the strength of relationships between theoretically selected baseline and time-varying covariates with patterns of consistency of condom use over time (baseline and 6 and 12 months). Based on reported condom use consistency at three points in time for each of two measures (global and partner specific), trajectories of condom use consistency were created using a subjective rule-based approach. All bivariate analyses utilized four condom use consistency trajectory groups as the outcome variable: stable use, stable nonuse, change to use, and change to nonuse.

Back to Top | Article Outline
Covariate Relationships

Based on extant evidence and consistent with current understanding about sexual behavior in adolescents, all covariates, with the exception of treatment group and age, were expected to be significantly associated with condom use trajectories. The near-equal number of girls in the treatment and control groups reflects the success of random assignment in the intervention trial from which these data were drawn and assures that it did not confound the other bivariate relationships examined in this analysis. Likewise, age of study participants at baseline did not confound other bivariate relationships in this study. Condom use at most recent intercourse generally declines with age (Centers for Disease Control and Prevention, 2010). Also consistently documented is that the younger the adolescent at sexual debut, the less apt he or she is to use any contraceptive method (Ford, Sohn, & Lepkowski, 2001; Mosher & Jones, 2010) In contrast, condom use is highest among those who are aged 16-17 years at sexual debut (Manlove et al., 2007; Mosher & Jones, 2010). In this analysis, rather than examining relationships between age, age at sexual debut, and condom use consistency reported at one point in time, we focused on age and longitudinal trajectories of condom use consistency. Age of the adolescent girls was not related to stability of condom use consistency over time.

Two hormonal use covariates were analyzed in relation to condom use consistency trajectory groups: baseline use and stability of use over time. Only for the partner-specific trajectory groups were relationships with these measures of hormonal use significant or trending toward significance. Among adolescents, the adoption of a hormonal method typically corresponds with a reduction in condom use (Bearinger & Resnick, 2003; Ott et al., 2002; Sieving, Bearinger, Resnick, Pettingell, & Skay, 2007). Findings for the partner-specific trajectory groups supported this dynamic; that is, girls in the stable use and change to use trajectory groups were least likely to report hormonal use at baseline, whereas girls in trajectory groups leading to nonuse of condoms over time were most likely to report hormonal use at baseline. Similarly, those who reported a decrease in condom use consistency use over time, in regard to their most recent sex partner, were most likely to report stable hormonal use over time.

Across adolescence, the stability of a relationship with a primary sex partner is a powerful influence on condom use decision making with that partner as well as with any concurrent sex partners (Fortenberry et al., 2002; Macaluso, Demand, Artz, & Hook, 2000; Manlove & Terry-Humen, 2007; Manning et al., 2009; Wiemann et al., 2009). Operationalized as a dichotomous variable, that is, girls who nominated the same primary sex partner at each of the three time points versus those who did not, our findings showed a significant relationship between primary sex partner stability and condom use consistency trajectory groups as determined by both the global and the partner-specific measures. In that this is one of the most reoccurring patterns in the literature, it might be expected that this covariate would be associated with both measures of condom use consistency. The significant relationship between stability of a primary partner and global condom use consistency trajectories suggests the powerful impact of primary partner stability on all aspects of decision making about condom use, beyond the primary sex partner.

One survey item was used to create two covariates relating to the number of vaginal sex partners of study participants. The first covariate tested the relationship between number of sex partners reported in the 6 months prior to baseline and the trajectory groups; the relationship was significant for the partner-specific but not the global trajectory group. Those in the stable use group had a higher number of partners at baseline than did those in the stable nonuse group. However, those in the change to nonuse group had a greater number of partners at baseline than did any other group. The second covariate using this item examined the impact of the stability of number of sex partners over time; that is, those who indicated only one partner in the 6 months prior to each of the survey time points were compared with those who indicated more than one partner at any of the study intervals. There are several possibilities as to why this stability covariate was nonsignificant for both trajectory groups. Perhaps, the actual difference in number of partners between the two groups (those with only one partner vs. those with more than one at each of the three time points) was not enough to "activate" the dynamic by which teenagers are more likely to change their condom use in response to having a greater number of partners. In the sample for this study, the mean number of sex partners in the 6 months prior to the baseline survey was 1.58, with an SD of 1.06, and although the range was 1-7 partners, only 9 of the 151 girls had ≥4 partners. Or, it is possible that during the 6-month time period assessed by this item, there were two or more sex partners in serial monogamy. In such a case, each successive partner relationship might have been perceived to be in an established relationship and, therefore, low risk (Fortenberry et al., 2002). Finally, in that condom use decisions vary by partner, it could be that with a greater number of partners, there is more opportunity to inconsistently use condoms (Manlove et al., 2007).

Back to Top | Article Outline
Global Versus Partner-Specific Measures

A comparison of significant covariate relationships with the two trajectory groups formed by classifying study participants according to responses to global and partner-specific measures provides an opportunity to interpret the relative merits of the two methods for measuring condom use consistency in an adolescent population. For the global trajectory groups, only one covariate relationship was significant: the stability of a primary sex partner over the three time points. In contrast, four of the covariates had relationships with the partner-specific trajectory groups that were significant or trended toward significance: hormonal use in the last 6 months (baseline), hormonal use stability (time varying), stability of primary sex partner (time varying), and number of sex partners in the last 6 months (baseline). Similar to the work of Wiebe, Guyatt, Weaver, Matijevic, and Sidwell (2003), who examined the "comparative responsiveness of generic versus specific quality-of-life instruments," our findings suggest greater capacity of the partner-specific measure to tap into a behavior, condom use consistency, hypothesized to be associated with variables expected to change that behavior, that is, stability of partner, number of sexual partners, and hormonal use.

The implications for these findings are threefold. First, the analysis provides insight regarding decisions around measurement strategies for assessing condom use consistency, whether for purposes of evaluating a health promotion intervention or determining country-level changes in adolescent sexual behavior. Particularly, when the need is to observe change in a behavior such as condom use, in which multiple factors are simultaneously influencing decisions and actions, there is great value in using the most responsive measure. Second, these implications are applicable to clinical practice in which brief, focused patient assessments should ideally pursue interview questions most apt to identify risk or threats to health. Based on these findings, adolescents should be queried on their condom use consistency with a specific partner rather than being asked about their use of condoms in general. Third, in this analysis, the simultaneous testing of these two measures, assessed over three time points, permitted the formation of parallel trajectory groups. Thus, the analysis was able to offer insight into the capacity of these measures for use in studies seeking to evaluate health promotion interventions focused on improving condom use consistency. In each of these applications, our findings suggest that focusing on partner-specific behavior, rather than a global measure of condom use consistency, will more accurately assess change and determine stability.

Back to Top | Article Outline
Limitations

Several aspects of the data set limited the analysis. As detailed previously, the small sample with data from only three time points, combined with limited within-person variability of condom use consistency over time for both global and partner-specific measures, narrowed analytic options. The total number of adolescent girls in the Prime Time randomized intervention trial was 253 youths; for this analysis, the full sample was narrowed to 185 based on inclusion criteria and completion of all items related to key variables. Finally, elimination of the unstable trajectory groups resulted in a final sample of 151. Consequently, bivariate analysis using a subjective rule-based approach to identifying condom use consistency trajectory groups as compared with a statistical approach may have constrained potential analyses and interpretation. A larger sample with more time points and great within-person variability in the outcome measures would have made this feasible.

Back to Top | Article Outline
Implications for Trajectory Approaches to Understanding Condom Use Consistency

One of the core aims of health promotion research is to capture the process of change-before, during, and after an intervention. Thus, methodological approaches for examining change must incorporate multiple waves of data and use measures sensitive to changes in the behavior being observed. Together, they are the sine qua non of intervention research. Given these two essential elements, the opportunity for modeling trajectories follows. In this analysis, the sample size, limited number of data collection points, and lack of within-person variability in condom use consistency meant that we could not utilize statistical modeling to form trajectory groups, yet the findings inform the next steps for examining change in condom use consistency. The significance of key covariates in this analysis affirms that reliable relationships found in studies employing a single outcome point apply when trajectories of condom use consistencies are the focus. Moreover, this study tests the responsiveness of global versus partner-specific measures of condom use consistency as the basis for forming trajectory groups, albeit derived with a subjective rule-based approach. The next step in confirming optimal measures of condom use consistency for understanding change would be the use of statistical modeling to form trajectory groups. Doing so would allow for moving beyond bivariate analysis to simultaneously test covariates in ways that account for measurement error. In other words, the statistical modeling approach adds a capacity for describing patterns of repeated observations, in our case, self-report of condom use consistency, while incorporating coefficients, a kind of latent variable, that together assess the fit of a model to the data (Fitzmaurice, Laird, & Ware, 2004). This level of sophistication in modeling self-reported behaviors in adolescent populations characterized by change will advance the capacity to understand processes by which adolescents adopt safer sexual behaviors. Furthermore, such a statistical modeling approach would improve capacity to examine the impact of health promotion interventions designed to reduce sexual risk outcomes such as pregnancy or sexually transmitted infections during adolescence.

Back to Top | Article Outline
Conclusion

Examining two longitudinal approaches for measuring condom use consistency in relation to a set of empirically related covariates in a multiwave sample of sexually active adolescent girls provides a unique opportunity for analyzing the relative merits of global versus partner-specific measures. Some covariates reveal significant associations with condom use trajectories, whereas others do not. However, significant bivariate relationships indicate overall greater responsiveness of the partner-specific measure of condom use consistency. Despite its limitations, this analysis sheds light on a measurement conundrum that has been an obstacle to comparing and contrasting indicators of condom use consistency during adolescence, when change epitomizes their development, their behavior, and their relationships.

Back to Top | Article Outline

References

Bearinger, L. H., & Resnick, M. D. (2003). STI and pregnancy protection in adolescents: A review and framework for action. Journal of Adolescent Health, 32, 340-349.

Brener, N. D., Billy, J. O. G., & Grady, W. R. (2003). Assessment of factors affecting the validity of self-reported health-risk behavior among adolescents: Evidence from the scientific literature. Journal of Adolescent Health, 33, 436-457.

Catania, J. A. (1999). A framework for conceptualizing reporting bias and its antecedents in interviews assessing human sexuality. Journal of Sex Research, 36, 25-38.

Catania, J. A., Gibson, D. R., Chitwood, D. D., & Coates, T. J. (1990). Methodological problems in AIDS behavioral research: Influences on measurement error and participation bias in studies of sexual behavior. Psychological Bulletin, 108, 339-362.

Centers for Disease Control and Prevention (CDC). (2010). Youth online: High school YRBS. United States 1991-2009 results. Retrieved on October 14, 2010, from http://apps.nccd.cdc.gov/youthonline/App/Results

Couper, M. P., Tourangeau, R., & Marvin, T. (2009). Taking the audio out of audio-CASI. Public Opinion Quarterly, 73, 281-303.

Coxon, A. (1999). Parallel accounts? Discrepancies between self-report (diary) and recall (questionnaire) measures of the same sexual behaviour. AIDS Care, 11, 221-234.

Crosby, R. A., DiClemente, R. J., Wingood, G. M., Salazar, L. F., Rose, E., Sales, J. M., et al. (2007). Oral contraceptive use may not preclude condom use: A study of non-pregnant African-American adolescent females. Sexually Transmitted Infections, 83(3), 216-218.

Crosby, R. A., Salazar, L. F., DiClemente, R. J., Yarber, W. L., Caliendo, A. M., & Staples-Horne, M. (2005). Accounting for failures may improve precision: Evidence supporting improved validity of self-reported condom use. Sexually Transmitted Diseases, 32, 513-515.

Crosby, R. A., Yarber, W. L., Sanders, S. A., & Graham, C. A. (2004). Condom use as a dependent variable: A brief commentary about classification of inconsistent users. AIDS and Behavior, 8(1), 99-103.

Fitzmaurice, G., Laird, N., & Ware, J. (2004). Applied Longitudinal Analysis. New York: Wiley.

Ford, K., Sohn, W., & Lepkowski, J. (2001). Characteristics of adolescents' sexual partners and their association with use of condoms and other contraceptive methods. Family Planning Perspectives, 33, 100-105.

Fortenberry, J. D., Wanzhu, T., Harezlak, J., Katz, B. P., & Orr, D. P. (2002). Condom use as a function of time in new and established adolescent sexual relationships. American Journal of Public Health, 92, 211-213.

Geary, W. C., Tchupo, J., Johnson, L., Cheta, C., & Nyama, T. (2003). Respondent perspectives on self-report measures of condom use. AIDS Education and Prevention, 15, 499-515.

Graham, C. A., Crosby, R. A., Sanders, S. A., & Yarber, W. L. (2005). Assessment of condom use in men and women. Annual Review of Sex Research, 16, 20-52.

Katz, B. P., Fortenberry, J. D., Wanzhu, T., Harezlak, J., & Orr, D. P. (2001). Sexual behavior among adolescent women at high risk for sexually transmitted infections. Sexually Transmitted Diseases, 28, 247-251.

Katz, B. P., Fortenberry, J. D., Zimet, G. D., Blythe, M. J., & Orr, D. P. (2001). Partner-specific relationship characteristics and condom use among young people with sexually transmitted diseases. Journal of Sex Research, 37, 69-75.

Macaluso, M., Demand, M. J., Artz, L. M., & Hook, E. W. (2000). Partner type and condom use. AIDS, 14, 537-546.

Manlove, J., Ryan, S., & Franzetta, K. (2007). Contraceptive use patterns across teens' sexual relationships: The role of relationships, partners, and sexual histories. Demography, 44, 603-621.

Manlove, J., & Terry-Humen, E. (2007). Contraceptive use patterns within females' first sexual relationships: The role of relationships, partners, and methods. Journal of Sex Research, 44, 3-16.

Manning, W. D., Flanigan, C. M., Giordano, P. C., & Longmore, M. A. (2009). Relationship dynamics and consistency of condom use among adolescents. Perspectives on Sexual and Reproductive Health, 41, 181-190.

McAuliffe, T. L., DiFranceisco, W., & Reed, B. R. (2007). Effects of question format and collection mode on the accuracy of retrospective surveys of health risk behavior: A comparison with daily sexual activity diaries. Health Psychology, 26, 60-67.

Mosher, W. D., & Jones, J. (2010). Use of contraception in the United States: 1982-2008. National Center for Health Statistics. Vital Health Statistics, 23(29), 1-38.

Noar, S. M., Cole, C., & Carlyle, K. (2006). Condom use measurement in 56 studies of sexual risk behavior: Review and recommendations. Archives of Sexual Behavior, 35, 327-345.

Ott, M., Adler, N. E., Millstein, S. G., Tschann, J. M., & Ellen, J. M. (2002). The trade-off between hormonal contraceptives and condoms among adolescents. Perspectives on Sexual and Reproductive Health, 34, 6-14.

Ramjee, G., Weber, A., & Morar, N. (1999). Recording sexual behavior: Comparison of recall questionnaires with a coital diary. Sexually Transmitted Diseases, 26, 374-380.

Rose, E., DiClemente, R. J., Wingood, G. M., Sales, J. M., Latham, T. P., Crosby, R. A., et al. (2009). The validity of teens' and young adults' self-reported condom use. Archives of Pediatrics and Adolescent Medicine, 163, 61-64.

Shrier, L. A., Shih, M., & Beardslee, W. R. (2005). Affect and sexual behavior in adolescents: A review of the literature and comparison of momentary sampling with diary and retrospective self-report methods of measurement. Pediatrics, 115, e573-e581.

Sieving, R. E., Bearinger, L. H., Resnick, M. D., Pettingell, S., & Skay, C. L. (2007). Adolescent dual method contraceptive use: Relevant attitudes, normative beliefs, and self-efficacy. Journal of Adolescent Health, 40, e15-e22.

Sieving, R. E., Hellerstedt, W., McNeely, C., Fee, R., Snyder, J., & Resnick, M. D. (2005). Reliability of self-reported contraceptive use and sexual behaviors among adolescent girls. Journal of Sex Research, 42, 159-166.

Sieving, R. E., McMorris, B. J., Beckman, K., Pettingell, S., Secor-Turner, M., Kugler, K., et al. (in press). Prime time: 12-month sexual health outcomes of a clinic-based intervention to prevent pregnancy risk behaviors. Journal of Adolescent Health.

Sieving, R. E., Resnick, M. D., Garwick, A. W., Bearinger, L. H., Beckman, K., Oliphant, J., et al. (in press). A clinic-based youth development approach to teen pregnancy prevention. American Journal of Health Behavior, 35(3), 346-358.

Sieving, R. E., & Shrier, L. A. (2009). Measuring adolescent health behaviors. In R. J. DiClemente, J. S. Santelli, & R. A. Crosby (Eds.), Adolescent health understanding and preventing risk behaviors (pp. 473-492). San Francisco: Jossey-Bass.

Weinhardt, L. S., Forsyth, A. D., Carey, M. P., Jaworski, B. C., & Durant, L. E. (1998). Reliability and validity of self-report measures of HIV-related sexual behavior: Progress since 1990 and recommendations for research and practice. Archives of Sexual Behavior, 27, 155-180.

Wiebe, S., Guyatt, G., Weaver, B., Matijevic, S., & Sidwell, C. (2003). Comparative responsiveness of generic and specific quality-of-life instruments. Journal of Clinical Epidemiology, 56(1), 52-60.

Wiemann, C. M., Chacko, M. R., Kozinetz, C. A., DiClemente, R., Smith, P. B., Velasquez, M. M., et al. (2009). Correlates of consistent condom use with main-new and main-old sexual partners. Journal of Adolescent Health, 45, 296-299.

Younge, S. N., Salazar, L. F., Crosby, R. F., DiClemente, R. J., Wingood, G. M., & Rose, E. (2008). Condom use at last sex as a proxy for other measures of condom use: Is it good enough? Adolescence, 43, 927-931.

Cited By:

This article has been cited 2 time(s).

Journal of Sexual Medicine
Understanding Human Papillomavirus Vaccination Intentions: Comparative Utility of the Theory of Reasoned Action and the Theory of Planned Behavior in Vaccine Target Age Women and Men
Fisher, WA; Kohut, T; Salisbury, CMA; Salvadori, MI
Journal of Sexual Medicine, 10(): 2455-2464.
10.1111/jsm.12211
CrossRef
Perspectives on Sexual and Reproductive Health
Life Experiences of Instability and Sexual Risk Behaviors Among High-Risk Adolescent Females
Secor-Turner, M; McMorris, B; Sieving, R; Bearinger, LH
Perspectives on Sexual and Reproductive Health, 45(2): 101-107.
10.1363/4510113
CrossRef
Back to Top | Article Outline
Keywords:

adolescents; condom use consistency; measurement; trajectories

© 2011 Lippincott Williams & Wilkins, Inc.

Login

Search for Similar Articles
You may search for similar articles that contain these same keywords or you may modify the keyword list to augment your search.